How AI answers decide which contractors to name, and how a small company gets in

AI assistants do not know who the good roofers are. When asked, they run a web search behind the scenes, read the top results, the map listings and the review sites, and summarize what those pages say. So a contractor gets named the way it gets found: by being on the pages the assistant reads, saying plainly what it does and where.

No trick replaces that, and a few settings quietly block it: a crawler rule that turns the assistant away at the door is the most common, and the one most owners do not know they have.

Where the answer gets its list

Each assistant works a little differently, but the shape is the same. ChatGPT's search, Google's AI Overviews and AI Mode, Perplexity and Claude all fetch live pages when a question is local or current.^1,2 What they fetch is, roughly in order:

1. The search results for the question, often the same first page a human would see, which puts you back on ordinary local search: the map pack and the first few organic results.
2. Business listings and review sites: Google Business Profile, Yelp, the BBB, Angi and HomeAdvisor, Nextdoor. The assistant reads your listing the way a person skims it: category, area, rating, number of reviews, recency.
3. Your own pages, if they can be fetched and if they say something quotable. A page that says "Roof repair and replacement in Schenectady, Niskayuna and Clifton Park since 2004, licensed and insured, 4.9 stars from 212 reviews" gives the assistant a sentence it can use. A page that says "Quality You Can Trust" over a photo gives it nothing.
4. Everyone else's pages that mention you: the local news story, the supplier's dealer locator, the chamber of commerce. These are why reputation work and listings matter more than any file you can add to your server.

An assistant names the companies it can read clearly, from sources it already trusts.

What a page needs to be quoted

Assistants prefer pages that answer a question in a sentence or two and then back it up. Four things, all cheap, all missing from most contractor sites:

• A plain statement of fact up top. Who you are, the trade, the towns, the license, the years, the reviews. In text, not in an image, in the first screen of the page.
• Structured data. A LocalBusiness (or the trade-specific type: RoofingContractor, HVACBusiness, Plumber, Electrician) block in the page's code with your name, address, phone, area served, hours and services. It is the machine-readable version of the sentence above, and both search engines and assistants read it.³
• Dates and an author. A page that says when it was written and by whom reads as a source. An undated page reads as a template.
• Questions and answers. A short Q&A on the page ("Do you work with insurance claims?" "Yes, and here is how it goes") gives the assistant the exact pairing a homeowner is asking for. Mark it up as FAQPage so the pairing is explicit; Google stopped showing FAQ rich results for most sites in 2023, but the markup is still read.⁴

This is also what the free audit's answer-engine check looks for: the structured data types present on your homepage, whether a business entity is declared, the heading structure, and whether the page states its town. Every page on this site, including this one, is built to those rules; so is every client site.

Let the right crawlers in

Assistants fetch pages with named crawlers, and websites can tell crawlers to stay out through a file called robots.txt. A lot of contractor sites block them without anyone deciding to: a hosting setting, a security plugin or a managed firewall rule that "blocks AI bots" by default. The result is that when an assistant goes to read your page, it is turned away, and it describes your competitor instead.

Our position for a local service business is simple: allow all of them, by name, in your own robots.txt, so a managed rule elsewhere cannot be read as a block. That is what this site does; you can read ours at /robots.txt. The short answer on blocking covers the case for blocking (there is one, for publishers; there is not one for a plumber).

llms.txt: what it is and what it is not

llms.txt is a proposed plain-text file at the root of a site that gives an assistant a short description of the business and a list of the pages worth reading, in a form a language model can take in at once.⁷ It was proposed in late 2024 and adoption is uneven: some assistants and tools read it, the big search-backed ones have not committed to it. It costs ten minutes and cannot hurt, so we put one on every site; ours is at /llms.txt. What it is not: a ranking signal, a way to get cited, or a replacement for a page that actually says what you do. The short answer has the format and an example.

What the audit checks, and what we are building

Today's free audit runs an answer-engine check on the homepage: which AI crawlers your robots.txt blocks, whether llms.txt exists, which structured-data types are present and whether a business entity is declared, the heading structure, the CMS, HTTP version and SSL, and a short simulation of how an assistant would describe the page. It reports those as a score and a list in section 4 of the PDF.

What we are adding, and will publish numbers from once it runs at scale: asking Google's AI the question a homeowner would ask about your trade in your town, recording which companies it names, and checking whether yours is one of them. The code for it is written; it is switched off until we are satisfied the answer is the same one a homeowner would get. When it runs, the share of contractors named per trade will appear in the data section, and the method page will say exactly how the question is asked.

What to do this week

1. Open yourdomain.com/robots.txt on your phone. If you see Disallow under any of the crawler names above, or your host has an "AI bot" toggle switched on, turn it off and name the crawlers as allowed.
2. Read your homepage's first screen as a stranger. Does it state the trade, the towns, the license and the years in plain text? If not, write that sentence and put it under the headline.
3. Check your Google Business Profile against the site: same name, same phone, same categories, same service area. Assistants cross-check; mismatches read as two different companies. The profile guide is the 20-minute version.
4. Ask ChatGPT and Google, out loud, "who is a good [trade] in [your town]?" Note who gets named. Those are the pages and listings the assistants are reading; look at what they say that yours does not.

Questions contractors ask about AI answers